Added vectorized aggregates

[kelindar]

This PR adds vectorized aggregates for Sum(), Avg(), Min() and Max(). Most of the magic is in bitmap and simd packages respectively.

Here are some benchmarks, sum is essentially a fully populated transaction (no filtering) and avg and max are calculated after an index (partial aggregates).

cpu: Intel(R) Core(TM) i7-9700K CPU @ 3.60GHz
BenchmarkCollection/sum-8                  89768             13822 ns/op              65 B/op          0 allocs/op
BenchmarkCollection/avg-8                  41806             28669 ns/op             124 B/op          0 allocs/op
BenchmarkCollection/max-8                  41257             28582 ns/op             131 B/op          0 allocs/op

Here what sum benchmark looked prior to this, ~40x improvement under optimal conditions (no filter) and ~1.4x improvement when aggregating over a filtered column which does not have contiguous blocks.

cpu: Intel(R) Core(TM) i7-9700K CPU @ 3.60GHz
BenchmarkCollection/sum-8                   2264            553207 ns/op            2322 B/op          1 allocs/op
BenchmarkCollection/avg-8                  29355             40862 ns/op             177 B/op          0 allocs/op

Note that it doesn't support ARM64 (e,g. Apple M1/M2) yet, as I haven't generated the relevant vectorized SVE assembly code.